in silico
Phase 1: In Silico Validation
- Objective
- Determine whether the proposed YAP/TAZ-Notch mechanotransduction cascade is computationally plausible in ovarian granulosa cells, and identify the critical stiffness threshold and inhibitor concentrations before committing to wet-lab experiments.
- Estimated cost
- €500-2000
- Estimated duration
- 4-6 weeks
- Success criteria
- Model prediction accuracy for YAP/TAZ target gene expression (CTGF, CYR61) in cryopreserved vs. fresh tissue · Pearson correlation ≥0.7 between simulated and observed fold-changes (from GEO datasets) · (Compare simulated fold-changes from Boolean model (steady-state) with DESeq2-derived log2 fold-changes from GSE123456)
- Predicted IC50 for verteporfin reducing YAP/TAZ nuclear ratio below 0.8 · IC50 between 2–8 µM (consistent with literature range 1–10 µM) · (Run model at 10 concentrations (0.1–20 µM) and fit a 4-parameter logistic curve)
- Predicted IC50 for DAPT reducing NICD intensity below baseline · IC50 between 5–15 µM (consistent with literature range 5–20 µM) · (Same approach as above, using NICD node activity as output)
- Go if
- Both predicted IC50 values fall within the literature ranges AND model predicts ≥50% reduction in YAP/TAZ nuclear ratio at 5 µM verteporfin and ≥60% reduction in NICD at 10 µM DAPT
- No-go if
- Model predicts no effect of stiffness on YAP/TAZ (nuclear ratio <1.2 at 8.0 kPa) OR IC50 values outside literature ranges (e.g., >20 µM for verteporfin)
- Pivot if
- Model predicts that YAP/TAZ and Notch act in parallel (no crosstalk) → redesign hypothesis to test independent pathways; OR if stiffness threshold is >5.0 kPa (unreachable by cryopreservation) → focus on other mechanotransducers (MRTF-A)
- Risks
- Boolean network model oversimplifies continuous dynamics (e.g., YAP/TAZ shuttling is graded, not binary)Probability: mediumMitigation: Use ordinary differential equations (ODE) model in COPASI or BioNetGen for critical parameters; validate with published time-course data (e.g., YAP/TAZ translocation kinetics from Dupont et al. 2011)
- GEO datasets may not include granulosa-cell-specific transcriptomes (bulk tissue data dilutes signal)Probability: mediumMitigation: Use single-cell RNA-seq data from ovarian cortex (e.g., GSE147745) if available; otherwise, perform deconvolution using CIBERSORTx to estimate granulosa cell contribution
- Model parameters (e.g., binding affinities) are unknown for ovarian granulosa cellsProbability: highMitigation: Use literature-derived parameters from other cell types (e.g., HEK293, MCF10A) and perform global sensitivity analysis to identify robust predictions; focus on qualitative trends (increase/decrease) rather than exact values